Inconsistencies between the ground and space environment, along with multiple disturbances, can seriously affect the control performance of coarse pointing assembly (CPA), thereby reducing the reliability of intersatellite laser communication links. To address accurate control and enhance the adaptability of the CPA, a self-learning metamodel disturbance observer-based composite controller is proposed. First, based on the practical operating environment, a CPA model is constructed with multiple disturbances. Second, with the state-space Kriging (SSK) metamodeling method, a self-learning metamodel disturbance observer is designed. Experimental data, partially known information, and disturbance estimation data are fully utilized to achieve high-precision estimation under uncertain operating environments. Based on the observer, a composite controller is designed to address time-varying precision constraints with barrier Lyapunov function (BLF). Finally, experiment results are introduced to demonstrate the effectiveness of the proposed method.